[Hive]Hive多分区操作

业务背景

mobile_log记录移动日志，现在需要将其保存到hive表中，将来可以按日期，小时进行统计，为此，需要建立一张具有日期、小时的hive分区表。

业务实现

hive分区表分为单分区表和多分区表，一个表可以拥有多个分区，每个分区都以文件夹的形式单独存放在表的文件目录下，详细可以参见Hive LanguageManual DDL

建立多分区表代码

drop table if exists pms.test_mobile_log;
create table pms.test_mobile_log 
(
    id          bigint,
    infomation  string
)
partitioned by (ds string, hour string)
row format delimited fields terminated by '\t'
lines terminated by '\n';

导入数据到多分区表中，实现方式有如下这些：

• 建表的时候，就插入数据，参考：

drop table if exists pms.test_mobile_log;
create table pms.test_mobile_log 
(
    id          bigint,
    infomation  string
)
partitioned by (ds string, hour string)
row format delimited fields terminated by '\t'
lines terminated by '\n';

insert overwrite table pms.test_mobile_log partition(ds='2015-05-26', hour='13') 
select 
    id,
    category_name
from category;

• 使用LOAD DATA方式导入数据，参考

load data inpath '/user/pms/workspace/ouyangyewei/temp2/category.txt' overwrite into table pms.test_mobile_log partition (ds='2015-05-26', hour='15');

• 新增分区时导入数据，参考：

alter table pms.test_mobile_log add partition (ds='2015-05-27', hour='14') location '/user/pms/workspace/ouyangyewei/temp2/category.txt';

实验结果

表结构

CREATE  TABLE pms.test_mobile_log(
  id bigint, 
  infomation string)
PARTITIONED BY ( 
  ds string, 
  hour string)
ROW FORMAT DELIMITED 
  FIELDS TERMINATED BY '\t' 
  LINES TERMINATED BY '\n' 
STORED AS INPUTFORMAT 
  'org.apache.hadoop.mapred.TextInputFormat' 
OUTPUTFORMAT 
  'org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat'
LOCATION
  'hdfs://yhd-jqhadoop2.int.yihaodian.com:8020/user/hive/pms/test_mobile'
TBLPROPERTIES (
  'numPartitions'='2', 
  'numFiles'='2', 
  'transient_lastDdlTime'='1432711793', 
  'numRows'='0', 
  'totalSize'='3517', 
  'rawDataSize'='0')

表分区

$hadoop fs -ls /user/hive/pms/test_mobile_log/ds=2015-05-26
Found 2 items
drwxr-xr-x   - pms supergroup  0 2015-05-27 13:53 /user/hive/pms/test_mobile_log/ds=2015-05-26/hour=13
drwxr-xr-x   - pms pms         0 2015-05-27 15:29 /user/hive/pms/test_mobile_log/ds=2015-05-26/hour=15